Atmospheric Oxalic Acid and SOA Production from Glyoxal: Results of Aqueous Photooxidation Experiments

ARTICLE IN PRESS

Atmospheric Environment 41 (2007) 7588–7602 www.elsevier.com/locate/atmosenv

Atmospheric oxalic acid and SOA production from glyoxal: Results of aqueous photooxidation experiments

Annmarie G. Carltona, Barbara J. Turpinb,Ã, Katye E. Altieric, Sybil Seitzingerc, Adam Reffd, Ho-Jin Lime, Barbara Ervensf

aASMD, ARL, NOAA, Mail Drop E-243-01, Research Triangle Park, NC 27711, USA bDepartment of Environmental Sciences, Rutgers University, 14 College Farm Road, New Brunswick, NJ 08901, USA cInstitute of Marine and Coastal Sciences, Rutgers University, Rutgers/NOAA CMER Program, 71 Dudley Road, New Brunswick, NJ 08901, USA dAMD, NERL, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA eDepartment of Environmental Engineering, Kyungpook National University, Daegu 702-701, Republic of Korea fDepartment of Atmospheric Science, Colorado State University, Fort Collins, CO 80523, USA

Received 12 February 2007; received in revised form 3 May 2007; accepted 16 May 2007

Abstract

Aqueous-phase photooxidation of glyoxal, a ubiquitous water-soluble gas-phase oxidation product of many compounds, is a potentially important global and regional source of oxalic acid and secondary organic aerosol (SOA). Reaction kinetics and product analysis are needed to validate and reﬁne current aqueous-phase mechanisms to facilitate prediction of in-cloud oxalic acid and SOA formation from glyoxal. In this work, aqueous-phase photochemical reactions of glyoxal and hydrogen peroxide were conducted at pH values typical of clouds and fogs (i.e., pH ¼ 4–5). Experimental time series concentrations were compared to values obtained using a published kinetic model and reaction rate constants from the literature. Experimental results demonstrate the formation of oxalic acid, as predicted by the published aqueous phase mechanism. However, the published mechanism did not reproduce the glyoxylic and oxalic acid concentration dynamics. Formic acid and larger multifunctional compounds, which were not previously predicted, were also formed. An expanded aqueous-phase oxidation mechanism for glyoxal is proposed that reasonably explains the concentration dynamics of formic and oxalic acids and includes larger multifunctional compounds. The coefﬁcient of determination for oxalic acid prediction was improved from 0.001 to 40.8 using the expanded mechanism. The model predicts that less than 1% of oxalic acid is formed through the glyoxylic acid pathway. This work supports the hypothesis that SOA forms through cloud processing of glyoxal and other water-soluble products of alkenes and aromatics of anthropogenic, biogenic and marine origin and provides reaction kinetics needed for oxalic acid prediction. r 2007 Elsevier Ltd. All rights reserved.

Keywords: Secondary organic aerosol; Aqueous-phase atmospheric chemistry; Glyoxal; Oxalic acid; Organic PM; Cloud processing

1. Introduction

The generally poor understanding of the sources ÃCorresponding author. Tel.: +1 732 932 9800x6219; fax: +1 732 932 8644. and formation of secondary organic particulate E-mail address: [email protected] (B.J. Turpin). matter (PM) is a major source of uncertainty in

1352-2310/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2007.05.035 ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7589 predictions of aerosol concentrations and properties pH values typical of clouds. Differences between that affect health, visibility and climate (EPA, 2004; aqueous- and gas-phase chemistry suggest that SOA IPCC, 2001; Kanakidou et al., 2005). There is formation from aldehydes is more favorable in the growing evidence suggesting that, like sulfate, aqueous phase than in the gas phase. The aqueous secondary organic aerosol (SOA) is formed through medium enables formation of new structures (i.e., aqueous-phase reactions in clouds, fogs and aero- gem diols) whose functional groups are oxidized sols (Blando and Turpin, 2000; Warneck, 2003; during reactions with dOH and other oxidants, Ervens et al., 2004; Crahan et al., 2004; Gelencser while the C–C bond structure is initially preserved. and Varga, 2005; Lim et al., 2005; Carlton et al., In contrast, in the gas phase, C–C bonds are usually 2006; Altieri et al., 2006). However, this formation broken yielding smaller, more volatile compounds pathway is poorly understood. As was the case for (e.g., GLY oxidizes to form volatile compounds, sulfate, model simulation is needed to evaluate the HO2, CO, HCHO, in the gas phase; Atkinson et al., regional and global importance of SOA formed as a 2006). result of aqueous-phase atmospheric chemistry. GLY is the gas-phase oxidation product of many This effort is hampered by the lack of kinetic data compounds of anthropogenic (Kleindienst et al., for recognized pathways and because many pro- 1999; Atkinson, 2000; Volkamer et al., 2001; ducts and pathways are unknown. Magneron et al., 2005; Volkamer et al., 2005), A large gap between measurements and model biogenic (Atkinson, 2000; Spaulding et al., 2003), predictions of organic PM was recently observed in and marine (Miller and Moran, 1997; Warneck, the free troposphere (Heald et al., 2005). This 2003) origin. It is found widely in the environment discrepancy might arise from atmospheric processes in the gas and aerosol phases and in cloud, fog and not yet parameterized in current models, such as in- dew water (Sempere and Kawamura, 1994; Matsu- cloud SOA formation. In-cloud SOA formation is moto et al., 2005). While GLY is present at likely to enhance organic PM concentrations in the concentrations (5–280 mM in cloud water; Munger free troposphere and organic aerosol concentrations et al., 1990) lower than SO2, at cloud relevant pH in locations affected by regional pollutant transport. the water solubility of GLY (effective Henry’s law 5 1 Predictions and experiments provide strong support constant, Heff43 10 M atm at 25 1C; Betterton for the following. Alkene and aromatic emissions and Hoffmann, 1988) is 3 orders of magnitude are oxidized in the interstitial spaces of clouds; the greater than that of SO2. (Cloud processing is an water-soluble products partition into cloud dro- important pathway for particulate sulfate formation plets, where they oxidize further forming low from SO2; Seinfeld and Pandis, 1998.) Also, GLY volatility compounds that remain at least in part has fast uptake by droplets (Schweitzer et al., 1998), in the particle phase after droplet evaporation, is observed in cloud water, and is highly reactive in forming SOA (Blando and Turpin, 2000; Warneck, the aqueous phase (Buxton et al., 1997). Therefore, 2003; Ervens et al., 2004; Crahan et al., 2004; since GLY is ubiquitous in the environment, can Gelencser and Varga, 2005; Lim et al., 2005; enter a cloud or fog droplet readily, and is predicted Carlton et al., 2006; Altieri et al., 2006). Recent to form low volatility compounds through aqueous- kinetic modeling supports in-cloud oxalic acid and phase photooxidation, SOA formation through thus SOA formation from glyoxal (GLY) and other cloud processing of GLY is likely. It is important gas-phase precursors (Warneck, 2003; Ervens et al., to note that low volatility products (e.g., glyoxylic 2004; Lim et al., 2005). Measured atmospheric and oxalic acids) are expected from aqueous-phase concentration dynamics suggest that GLY is an in- GLY oxidation but gas-phase oxidation produces cloud precursor for carboxylic acids (Chebbi and high volatility compounds (e.g., HO2, CO, HCHO) Carlier, 1996) that likely contribute to SOA due to not expected to contribute to SOA directly (Ervens their low volatility (e.g., glyoxylic and oxalic acids). et al., 2004). Other similar water-soluble organics Batch photochemical experiments support the in- are found in clouds (Kawamura et al., 1996a, b) and cloud SOA hypothesis through product analysis are likely to contribute to in-cloud SOA formation that demonstrates low volatility product formation as well (e.g., methylglyoxal and glycolaldehyde; from GLY (e.g., glyoxylic acid, Buxton et al., 1997) Warneck, 2003; Ervens et al., 2004; Lim et al., and pyruvic acid (e.g., glyoxylic and oxalic acids, 2005). Carlton et al., 2006, and larger oligomeric com- Dicarboxylic acids are similarly ubiquitous in the pounds, Altieri et al., 2006; Guzman et al., 2006)at atmosphere and oxalic acid is the most abundant ARTICLE IN PRESS 7590 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 dicarboxylic acid (Kawamura et al., 1996a, b; Yu (OH) CHCH(OH) a 2 2 et al., 2005). Primary sources of oxalic acid exist (glyoxal-hydrated) ·OH c (e.g., fossil fuel combustion), however they are b k = 3E10 ·OH k = 5E3 insufficient to support measured ambient concen- ·OH k = 1.1E8 H O 2 2 large trations (Yu et al., 2005). There is growing evidence ·OH H2O2 multifunctional (OH) CHCOOH HCOOH from atmospheric observations that oxalic acid is a 2 compounds (glyoxylic acid-hydrated) (formic acid) product of cloud processing (Kawamura and Gagosian, 1987; Kawamura and Usukura, 1993; ·OH ·OH ·OH Chebbi and Carlier, 1996; Crahan et al., 2004; Yu et ·OH HOOCCOOH CO al., 2005; Sorooshian et al., 2006; Heald et al., (oxalic acid) 2 2006). For example, Crahan et al. (2004) measured in-cloud and below-cloud oxalate in the coastal Fig. 1. Aqueous-phase glyoxal oxidation pathways. Glyoxal, marine atmosphere and found that, as for sulfate glyoxylic acid and glyoxylate are predominantly hydrated in solution (Ervens et al., 2003b). The initial mechanism (pathway with a known in-cloud production mechanism, the ‘‘b’’, shaded) is adapted from Ervens et al. (2004). Pathways ‘‘a’’ in-cloud concentration was approximately three and ‘‘c’’ are supported by experimental evidence contained times the below-cloud concentration and the size herein. distributions of sulfate-containing and oxalate- containing particles were similar. Sorooshian et al. Table 1 (2006) compared measured and modeled in-cloud Glyoxal experimental design oxalate concentrations from the International Con- sortium for Atmospheric Research on Transport Initial glyoxal conc. 2 mM and Transformation (ICARTT) study and con- Initial H2O2 conc. 10 mM Number of experiments 2 cluded that cloud processing was the major source pH 4.1–4.8 of atmospheric oxalate. They observed that (1) the Temperature 2571 1C detection frequency and particulate oxalate concen- Experiment GLY+UV+H2O2 tration in cloud-free air parcels were significantly UV control GLY+H2O2 lower than for samples collected in-cloud, (2) the H2O2 control GLY+UV Organic control UV+H O highest oxalate concentrations (aerosol and droplet 2 2 residuals) were observed in clouds influenced by Note: GLY ¼ glyoxal. anthropogenic plumes, and (3) sulfate and oxalate were correlated though they are not linked by hypothesis that SOA forms through cloud proces- production chemistry. sing. Current aqueous-phase models that predict in- cloud oxalic acid formation from GLY assume 2. Methods GLY is oxidized to glyoxylic acid and subsequently to oxalic acid (shaded pathway, Fig. 1). However, 2.1. Batch reactions organic product analysis in previous GLY aqueous oxidation experiments was limited to glyoxylic acid Batch photochemical aqueous reactions of (Buxton et al., 1997); oxalic acid formation was not GLY and hydrogen peroxide were conducted as confirmed, nor was the potential formation of other described previously in detail (Carlton et al., 2006). low volatility products investigated. The controlled Experimental conditions are listed in Table 1.UV laboratory experiments and product analysis pre- photolysis of hydrogen peroxide (H2O2) provided sented below demonstrate the formation of oxalic a source of dOH for GLY oxidation. The UV acid and other low volatility products from aqueous source was a low-pressure monochromatic (254 nm) photooxidation of GLY at pH values typical of mercury lamp (Heraeus Noblelight, Inc. Duluth, clouds. Detailed product analysis was used to GA) in a quartz immersion well in the center identify major mechanistic revisions that enabled of a 1 L borosilicate reaction vessel (ACE Glass accurate prediction of oxalic acid formation in the Inc., Vineland, NJ). For each experiment reaction vessel. The expanded reaction mechanism (GLY+UV+H2O2), three types of control experi- can be used to refine cloud chemistry models. The ments were performed: (1) GLY+H2O2 without formation of low volatility species through aqueous UV, (2) GLY+UV without H2O2, and (3) photooxidation of GLY provides support to the H2O2+UV without GLY. ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7591

Reaction solutions were prepared in 1 L volu- 2.2. Analytical procedures metric flasks and then poured into the reaction vessel. Solutions were continuously mixed, main- 2.2.1. High-performance liquid chromatography tained at constant temperature and ambient UV (HPLC)– UV/Vis analysis of organic acids penetration was minimized. All experiments began Organic acid analysis is described in detail in the with oxygen-saturated solutions. Samples for H2O2 supporting information of Carlton et al. (2006). and organic analysis were taken as follows. The first Briefly, all standards and samples were analyzed in sample was taken directly from the volumetric flask. triplicate for carboxylic acids by HPLC with UV The second sample was taken immediately after the absorbance detection at 205 nm (Beckman Coulter, solution transfer to the reaction vessel, with a third System Gold, Fullerton, CA). The HPLC employed sample taken after 5 min. Samples were then an Alltech, organic acid ion exclusion column (OA taken at 10 min intervals for the 1 h experiment 2000) with the corresponding guard column. The (Experiment 1) and 30 min intervals, with an extra stationary phase was sulfonated polystyrene divi- sample during the first hour, for the 5+ h nylbenzene and is specifically designed to retain experiment (Experiment 2). H2O2 in the experiment only compounds with organic acid and/or alcohol and control samples was destroyed through the functional groups (www.alltechweb.com). Com- 1 addition of catalase (H2O2-H2O) (0.25 mL1mL pounds containing multiple functional groups are of sample) immediately after sampling (Stefan expected to interact with the column multiple times et al., 1996). Samples were stored frozen until and generate broad peaks, according to column analysis. specifications. Compounds without these function- Initial GLY concentrations in the experiments alities, should they be present, are not retained by were greater than those typically found in cloud the column and elute immediately. In the chroma- and fog droplets (Matsumoto et al., 2005), but togram unretained products are contained within concentrations this high have occasionally been the void volume, the initial peak associated with observed in the ambient atmosphere (Munger et al., sample injection. The mobile phase was H2SO4 1995). It is worth noting also that cloud droplet (pH ¼ 2.3); the flow rate was 0.7 mL min1 and the evaporation leaves aerosol particles with very column temperature was maintained at 45 1C. The concentrated aqueous solutions (i.e., exceeding mean absorbance (71 standard deviation of tripli- concentrations used in these experiments). End cate analyses) was used for quantitation. Multi- products and reaction rate constants for GLY variate calibration was used to quantify organic experiments are not expected to be concentration acids analyzed by HPLC–UV. Partial least squares dependent below 1 M (above 1 M, GLY poly- (PLS) regression models (Martens and Naes, 1989) merizes) (Whipple, 1970; Kunen et al., 1983; were built from concentrations and chromatograms Hastings et al., 2005). of calibration standards spanning the range of Reaction solutions were prepared with excess sample concentrations and applied to experimental oxidant in order to have pseudo-first-order kinetics samples using Statistical Analysis System software with regard to GLY. However, initial oxidant levels (SAS, V8.2, Cary, NC). A set of 25 calibration were limited by the need to keep quenching times mixtures with orthogonal concentrations (Brereton, (i.e., time to completely destroy H2O2 in the 1997) was used to quantify glyoxylic and oxalic samples) short and maintain laboratory safety acid. Formic and acetic acids were calibrated with (e.g., by using H2O2 concentrations and H2O2-to- single component standards at five concentration catalase ratios that have been previously employed; values. GLY and catalase were analyzed alone and Stefan et al., 1996). The quenching time was less added to 12% of the calibration standards. Neither than 2 min for the first sample and sharply was detected in the chromatograms, and chromato- decreased with time as the H2O2 concentration in grams of standards with and without GLY and the reaction solutions decreased. Initial modeling catalase were indistinguishable. (GLY was detected using state-of-the-art aqueous-phase mechanisms by electrospray ionization–mass spectrometry, for GLY (Ervens et al., 2004; Lim et al., 2005) ESI–MS; see below.) Ten percent of the mixture suggested that the initial conditions of Table 1 standards were re-analyzed and independent single would yield intermediate and end product concen- component standards were analyzed to assess trations that would be sufficiently above detection analytical accuracy. Recoveries for individual acids limits on the time scale of the experiments. were calculated by placing standards in the reaction ARTICLE IN PRESS 7592 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 vessel and sampling the solution as for experimental a capillary voltage of 3000 V. Nitrogen was the samples. Ten percent of experiment samples drying gas (350 1C, 24 psig, 10 L min1). The unit were collected in duplicate to determine method mass resolution spectra were recorded on Agilent precision. software (Chemstation version A.07.01) and ex- ported to Access and Excel (Microsoft, Inc.) for 2.2.2. Hydrogen peroxide (H2O2) statistical analysis and interpretation. H2O2 in the time series samples was quantified by The ESI–MS uses a soft ionization process that the triiodide method (Klassen et al., 1994) within 1 h does not fragment compounds at the low voltages of sampling. Solutions were prepared according to used and provides molecular weight information Allen et al. (1952) and were not stored longer than 1 with unit mass resolution. The positive ionization month. Calibration was performed with 5 H2O2 mode protonates compounds with basic functional concentration values and a blank of milli-Q water groups (e.g. methyl, carbonyl) while the negative (18 MO). All samples were analyzed in triplicate, ionization mode deprotonates compounds with one calibration standard was re-analyzed after acidic functional groups (e.g. carboxylic acids). sample analysis and one independent standard was Single and mixed standards of GLY, glyoxylic acid analyzed during calibration. Method detection and oxalic acid (plus H2O2 at a 1:2 ratio and limits were determined from analysis of 8 indepen- catalase (0.5%)) were analyzed using the same dent blank solutions, 8 times. instrument conditions as the experimental samples (see Supporting Information, Figure S-1). Oxalic 2.2.3. Photon flux acid (m/z 89) and glyoxylic acid (m/z 73) were The lamp intensity was measured before and after detected as monomers in the negative mode as the GLY experiments using iodide–iodate actino- would be expected for carboxylic acids (Figure S-1). metry (Rahn et al., 2003). Briefly, a 1 L actinometer GLY was detected in the positive mode as a dimer solution of 0.6 M KI, 0.1 M KIO3 and 0.01 M (m/z 117; twice molecular weight plus one). Na2B4O7 10H2O (borax) at pH 9.25 is prepared Aldehyde dimerization is common during ESI immediately prior to the photon flux measurement. analysis, in particular for GLY (Hastings et al., The reactor is filled with actinometer solution and 2005; Loeffler et al., 2006). In addition to the GLY exposed to the mercury lamp. Samples are collected dimer ion, a second qualifying ion (m/z 131) was as quickly as possible (every 40 s). The absorbance detected for GLY. The composition of this ion is of the solution is measured immediately with a unknown, but it appears in concert with the main spectrometer at 476 nm. An absorbance blank of GLY ion, and linearly increases in ion abundance milli-Q water was subtracted from the sample when GLY concentration increases. The qualifier absorbances for photon flux calculations, described ion was used for identification purposes only. below. The intensity of the irradiation source received by solutions in the reaction vessel was calculated using the method described by Murov (1973). 2.2.5. Dissolved organic carbon (DOC), pH, dissolved oxygen and temperature 2.2.4. ESI– MS Samples were analyzed for bulk DOC using a Selected samples were analyzed by ESI/MS (HP- Shimadzu 5000A high-temperature combustion Agilent 1100) as described previously (Seitzinger et analyzer (Sharp et al., 1993). The initial and final al., 2005; Altieri et al., 2006). Qualitative ESI results pH (Oakton Instruments Vernon Hills, IL) and are presented here. An autosampler injected sample dissolved oxygen (DO; YSI Inc., Yellowsprings, solutions (20 mL) from individual vials into a liquid OH) concentrations were also measured. The pH chromatography (LC) system, which introduces the meter was calibrated at pH ¼ 4, 7, and 10; verifica- sample into the ESI source region. All samples were tion standards (pH ¼ 2, H2SO4;pH¼ 3, HClO4) analyzed with no LC column attached. The mobile were analyzed each day of use. DO readings were phase was 60:40 v/v 100% methanol and 0.05% verified daily using O2-saturated solutions with formic acid in deionized water with a flow rate of known saturation values. Temperature was mea- 0.220 mL min1. Samples were analyzed in the sured throughout the experiment with an alcohol negative and positive mode over the mass range thermometer that was verified 72 1Cattwo 50–1000 amu with a fragmentor voltage of 40 V and temperatures. ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7593

2.3. Modeling glyoxylic acid (m/z 73) and oxalic acid (m/z 89) were detected in the negative mode. Note that Carboxylic acid concentration time profiles were formic acid has a molecular weight below the predicted using a commercially available differential ESI–MS instrument detection limit (m/z 50). The equation solver (FACSIMILE; AEA Technology, ESI–MS and HPLC control experiment results are Oxfordshire, UK). Initial kinetic modeling was provided in Supporting Information. based on the aqueous-phase mechanism of Ervens et al. (2004), which is summarized in Fig. 1 (shaded 3.2. Photon flux pathway) and reactions 1–16 in Table 2. The corollary anion reactions and the acid/base equili- The calculated mean intensities from photon bria are not depicted in Fig. 1 though they occur fluence measurements conducted before and after (Stefan et al., 1996; Ervens et al., 2003a, b) and were d the GLY experiments agreed within one standard included in the model. The concentration of OH d deviation. Hence, these experiments demonstrated was not explicitly measured; the OH concentration that the received lamp intensity was constant during time series was predicted (Table 2, reactions 1–4; the experiments and the H O photolysis reaction Liao and Gurol, 1995) and the accuracy of these 2 2 rates (reaction 1 in Table 2) were constant across the predictions was verified with H O measurements as 2 2 experiments. The photolysis reaction rate constant shown below. After examination of the concentra- (k in Table 2) was determined as described below. tion dynamics of products, an expanded reaction 1 mechanism was proposed (all pathways, Fig. 1; all reactions, Table 2). The expanded mechanism, 3.3. Hydrogen peroxide (H2O2) measured concentrations and the differential equa- d tion solver were used to fit unknown reaction rate The photolytic decomposition of H2O2 to OH in constants and predict the formation of newly pure water is well understood (Liao and Gurol, identified products. 1995; Stefan et al., 1996) and is described in the first 4 reactions of Table 2.H2O2 concentrations from 3. Quality control results H2O2+UV control experiments (N ¼ 2) and model predictions agree well (Fig. 2) providing confidence 3.1. Organic measurements that concentrations of dOH are described well in the experiments. (Note that reaction rate constants The HPLC analysis was used for identification k2–k4 are known and the photolysis rate (k1) and quantification of compounds, while the depends on photon fluence from the lamp and was ESI–MS analysis was used only for identification a fitted parameter.) While experimental and control of compounds. The PLS and single-acid calibrations solutions were prepared with 10 mM H2O2,anH2O2 described more than 96% of the variance in concentration of 8 mM was used to initialize the concentration of each carboxylic acid in the com- model simulation because H2O2 concentrations in plex mixture standards analyzed by the HPLC. the GLY+H2O2 control experiments were stable at Quality control measures for organic acids are given 8 mM (see Supporting Information, Figure S-2). d in Table 3. Recoveries were high, and the lowest This is reasonable because H2O2 photolyzes to OH recovery (80%) was obtained for the most volatile at wavelengths (l) present in ambient light and the compound (formic acid) as expected. (Formic acid solution was exposed to ambient light for the 3- concentrations were corrected for recoveries.) Meth- 4 min required for solution preparation and transfer od detection limits were determined from analysis of into the shielded reaction vessel. H2O2 was not eight ‘‘organic control’’ (i.e., H2O2 and UV) samples detected in the GLY+UV control samples. H2O2 (Greenberg et al., 1991). Method precision is the photodecomposition experiments (H2O2+UV) pooled coefficient of variation of concentrations have also been performed at an initial H2O2 measured in duplicate samples. Accuracy was concentration of 20 mM (i.e., as part of the pyruvic calculated as the percent difference between the acid experiments; Carlton et al., 2006). The model actual and measured concentrations of independent also successfully reproduced these measurements. standards of individual compounds not used in the These experiments provided a photolysis reaction calibration models. In the ESI–MS, GLY (m/z 117, rate constant of 1.0(70.2) 104 s1, which was 131) was detected in the positive mode, while used in model simulations (Table 2, reaction 1) and ARTICLE IN PRESS 7594 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602

Table 2 Glyoxal oxidation mechanism (initial: reactions 1–16; expanded: reactions 1–28)

Reaction Rate constant (M1 s1) Reference Estimated, measured or ﬁtted

1 a 1 H2O2+hu-2OH 1.0E4(s ) Liao and Gurol (1995) m b 2 OH+H2O2-HO2+H2O 2.7E+07, 2.7E+10 Liao and Gurol (1995) m, f 3 HO2+H2O2-OH+H2O+O2 3.7 Liao and Gurol (1995) m 4 HO2+HO2-H2O2+O2 8.3E+05 Liao and Gurol (1995) m 5 GLY+OH-GLYAC+HO2 1.1E+09/1.1E+08 Buxton et al. (1997) m, f 6 GLYAC+OH- 3.6E+08 Ervens et al. (2003b) m

OXLAC+HO2+H2O 7 GLYAC+OH- 2.6E+09 Ervens et al. (2003b) m OXLAC +HO2+H2O 8 OXLAC+2OH-2CO2+2H2O 1.4E+06 Ervens et al. (2003b) m 9 OXLAC+OH- 1.9E+08/2.2E+08 Ervens et al. (2003b) mc CO2+CO2 +2H2O 10 OXLAC2+OH- 1.6E+08 Ervens et al. (2003b) m CO2+CO2 +OH 11 CO2 +O2-O2 +CO2 2.4E+09 Buxton et al. (1988), Warneck m (2003) + 12 H2O ¼ H +OH Keq ¼ 1.0E14, ka ¼ 1.4E11 Warneck (1999), Lelieveld and m Crutzen (1991) 13 HO2 ¼ Hp1+O2 Keq ¼ 1.6E5, ka ¼ 5.0E10 Warneck (1999, Ervens et al. m, e (2003a) + 14 GLYAC ¼ H +GLYAC Keq ¼ 3.47E4, ka ¼ 2.0E10 Warneck (2003), Ervens et al. m, e (2003a) + 15 OXLAC ¼ H +OXLAC Keq ¼ 5.67E2, ka ¼ 5.0E10 Warneck (2003), Ervens et al. m, e (2003a) + 2 16 OXLAC ¼ H +OXLAC Keq ¼ 5.42E5, ka ¼ 5.0E10 Warneck (2003), Ervens et al. m, e (2003a) 17 GLY+2OH- 5.0E+03 n/a f

HCO2H+HCO2H 18 HCO2H+OH- 1.3E+08 Ervens et al. (2003b) m CO2+HO2+H2O 19 HCO2 +OH-CO2 +H2O 3.2E+09 Ervens et al. (2003b) m + 20 HCO2H ¼ HCO2 +H Keq ¼ 1.77E4, ka ¼ 5.0E10 Ervens et al. (2003b) m, e 21 GLY+OH-Products 3.E+10 n/a fd 22 Products+OH-OXLAC 3.E+10 n/a fd 23 Products+OH-GLYAC 1.E+09 n/a fd

24 GLY+H2O2- 1 n/a f HCO2H+HCO2H e 25 GLYAC+H2O2- 0.9 n/a f HCO2H+CO2+H2O e 26 HCO2H+H2O2-CO2+H2O 0.2 n/a f e 27 OXLAC+H2O2-2CO2 0.11 n/a f e 28 OXLAC -2CO2 1.5E04 n/a f

Notes: Reactions 1–16 (italicized), used in initial mechanism, Reactions 17–28 added for expanded mechanism. GLY ¼ glyoxal, GLYAC ¼ glyoxylic acid, OXLAC ¼ oxalic acid, OH ¼ dOH; m ¼ measured, e ¼ estimated, f ¼ fitted; dissociation rate constants (kd) are calculated from the equilibrium constant (Keq; i.e., kd ¼ Keq ka). a k1 is a fitted parameter with observations fit to the Liao and Gurol (1995) parameterization. b Modeling was performed with k2 ¼ 2.7E+10, which likely represents the net rate of several reactions: OH+H2O2-HO2+H2O; OH+HO2-H2O+O2 (1.1E10, Elliot and Buxton (1992)); OH+RO2-products (E10). Note that H2O2 was insensitive to the choice of k2 (measured vs. fitted). cThis reaction rate constant is within the uncertainty of the measured value present by Ervens et al. (2003b) (k ¼ 1.9(70.6)E+08). dThese simplified reactions are surrogates for unknown formation processes. e These fitted reaction rate constants for H2O2 reaction with carboxylic acids are reasonable and close in value to the measured reaction 1 rate constant for pyruvic acid+H2O2 (0.11 M s ; Stefan and Bolton (1999)). ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7595

Table 3 Carboxylic acid quality control measures

Formic acid Glyoxylic acid Oxalic acid

Recovery 80% (N ¼ 2) 90% (N ¼ 2) 100% (N ¼ 2) Method detection limit 0.07 mM (N ¼ 8) 0.12 mM (N ¼ 8) 0.06 mM (N ¼ 8) Method precision Pooled C.V. ¼ 17%, N ¼ 5 Accuracy 15% (N ¼ 2) 10% (N ¼ 3) 5% (N ¼ 3)

10 Positive mode Negative mode Control Experiment 1 100000 40000 9 Control Experiment 2 117 8 Model Prediction 80000 30000 7 60000 131 6 20000 40000

5 10000 0 min. 4 20000 3 0 0 Concentration (mM) 0 200 400 600 800 0 200 400 600 800 2 2 O

2 60000 40000

H 1 149 0 55000 30000 0 50 100 150 50000 Time (min) 20000

10000 10000 9 min. Fig. 2. Hydrogen peroxide measurements and model predictions. d 1 The photolysis rate constant for H2O2-2 OH, (1.0E04 s ), 0 0 was a ﬁtted parameter (see Table 2, reaction 1). Note that the 0 200 400 600 800 0 200 400 600 800 60000 H2O2 model concentration at t ¼ 0 was considered to be equal to 40000 50000 the H2O2 concentration observed throughout the GLY+H2O2 Ion abundance 40000 173 30000 149 control experiment (Figure S-2). This accounts for H2O2 30000 119 photolysis due to ambient UV during solution preparation. 10000 20000 205

10000 89 32 min.

0 0 is similar to values obtained by others conducting 0 200 400 600 800 0 200 400 600 800 similar experiments (Herrmann et al., 1995). 60000 40000 89 50000 30000 4. Experimental results 40000 30000 20000 20000 201 4.1. Concentration dynamics 10000 151 min 10000 313 0 0 0 200 400 600 800 0 200 400 600 800 GLY (m/z 117, 131) was present in the 0 min m/z m/z sample and was completely absent by 9 min as demonstrated in the ESI mass spectra (Fig. 3). This Fig. 3. ESI mass spectra of selected samples from experiment 2 decrease in GLY coincided with the very rapid with increasing reaction time. Note change in y-axis scale for the m z formation of formic acid (Fig. 4; HPLC retention first positive mode spectrum. Glyoxal ( / 117, 131) is clearly present in the 0 min sample. Oxalic acid (m/z 89) is clearly present time (RT) 3.3 min). The formic acid concentration in the 151 min spectrum. Molecular level identification of other reached a maximum within the first 2 min of the products will require higher-resolution ESI–MS or other com- experiment (Fig. 4); it was not a predicted product. plementary analytical techniques. Glyoxylic acid, an expected product (Fig. 1), has a maximum HPLC absorbance at 9 min, however was evident in both the HPLC chromatograms and observations never exceeded detection limits (Figs. 4 the ESI mass spectra. By 9 min the positive mode and 5). Total DOC decreased by approximately a ESI–MS spectra showed numerous peaks clustered factor of 10 within the first hour. between m/z 80–200 (Fig. 3). At 30 min, there still In addition to the compounds predicted in the were higher molecular weight peaks in the positive initial reaction scheme (shaded pathway, Fig. 1), the mode spectra and, in addition, there were higher formation of larger molecular weight compounds molecular weight peaks in the negative mode ARTICLE IN PRESS 7596 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 Arbitrary Absorbance (mAU)

Retention Time (min)

Fig. 4. HPLC chromatograms of selected samples from experiment 2. Oxalic acid plateaus at 50–150 min and is the only product observed in the chromatograms after the 146 min sample. Formic acid appears quickly and is no longer present after the 32 min sample. The unresolved carbon maximum occurs at 30 min. Note: mAU ¼ milli-absorbance units; *indicates the small glyoxylic acid presence in the 32 min sample. Note that the y-axis units are arbitrary absorbance units. Note that unretained species are included in the void volume peak. spectra. After 30 min, the complexity in the area decreased. Column characteristics indicate that spectra of both positive and negative modes the shift to a longer retention time is most likely dissipated. (Note: the simple mass spectra of associated with an increase in compound pKa,or controls and mixture standards containing all molecular size. The decrease in area strongly expected products, shown and discussed in Support- suggests a decrease in concentration, however this ing Information, provides strong evidence that the is not explicitly known since the analytical sensitiv- ESI–MS ‘‘complexity’’ is not an instrumental ity could change with the changing product mix. artifact.) A feature with the same temporal evolu- A dramatic increase in the oxalic acid concentration tion was found in the HPLC chromatograms occurred after 30 min (Figs. 3–5), as the ESI–MS (Fig. 4). The HPLC chromatograms exhibited a ‘‘complexity’’ dissipated and the broad peak of broad peak with a long retention time whose area unresolved carbon in HPLC chromatograms de- reached a maximum at 30 min (Fig. 4). The broad creased. Glyoxylic acid concentrations are insufﬁ- peak suggests the formation of larger, multifunc- cient to account for this increase in oxalic acid. tional products, and the long retention time suggests Concentration dynamics suggest that the degrada- relatively high pKa values (typical of alcohols). tion of larger multifunctional compounds is some- After 30 min, the retention time of this peak how responsible. The oxalic acid concentration increased (shifted from 6.4 to 8 min) and the peak reached a plateau at 50–150 min (HPLC, ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7597

functional compounds and rapid formic acid 3.0 formation and degradation were not predicted by 2.5 Exp.1 the original mechanism, and the current experi- 2.0 Exp.2 ments suggest that glyoxylic acid oxidation is not 1.5 the predominant oxalic acid formation pathway. 1.0 Published atmospheric aqueous-phase models predict oxalic acid production from GLY via glyoxylic 0.5 MDL 3σ acid (pathway b; Fig. 1). While glyoxylic acid Concentration (mM) 0.0 oxidation could explain the initial formation of 0 20 40 60 80 100 Time (min) oxalic acid, glyoxylic acid concentrations are insufﬁcient to account for the substantial oxalic acid growth after 30 min. Oxalic acid forms rapidly 0.4 Exp.1 as the broad HPLC peak suggestive of larger multifunctional compounds decreases and as the 0.3 Exp.2 ESI mass spectral complexity decreases, suggesting 0.2 that oxalic acid forms primarily through the degradation of a class of larger multifunctional 0.1 MDL 3σ compounds (pathway a; Fig. 1). The model using the initial mechanism (Table 2, Concentration (mM) 0.0 reactions 1–16) failed to reproduce the concentra- 0 50 100 150 200 250 300 350 Time (min) tions of measured species or their temporal dynamics in the experiments (Fig. 6). Over-prediction of glyoxylic acid concentrations has been 0.4 Exp.1 reported for previous GLY photooxidation experi- Exp.2 ments as well (Buxton et al., 1997). 0.3 Based on the experimental results, the original mechanism (shaded pathway, Fig. 1) was expanded 0.2 to include the formation and degradation of formic MDL 3σ 0.1 acid and larger molecular weight compounds (all pathways, Fig. 1). Table 2 presents the reactions Concentration (mM) 0.0 and ﬁtted rate constants used in the expanded 0 50 100 150 200 250 300 350 mechanism model. Reactions were added for Time (min) formic acid formation, dissociation and oxidation (reactions 17–20, 24–26). The rapid formation Fig. 5. Formic acid (a), oxalic acid (b), and glyoxylic acid (c) concentration in glyoxal experiments. The experiments had good of formic acid very early in the time series repeatability; carboxylic acid concentrations across experiments suggests it is a direct GLY oxidation product; its were within measurement uncertainties. MDL is method detection limit. 1.5

RT ¼ 1.5 min; ESI–MS, m/z 89 negative mode) 1.0 Predicted glyoxylic Predicted oxalic (Figs. 3–5). After 150 min, oxalic acid was the only Oxalic Acid, Exp. 1, 2 peak present in the chromatograms and dominated 0.5 the ESI mass spectra as its concentration gradually decreased. Concentration (mM) 0.0 4.2. Modeling organic acid formation 0 100 200 300 Time (min) These experiments demonstrate that while oxalic Fig. 6. Measured and predicted concentrations of oxalic acid acid does form from aqueous-phase photooxidation using initial mechanism (shaded pathway, Fig. 1; Table 2, of GLY, as expected, the mechanism is more reactions 1–16). Note that observed glyoxylic acid concentrations complex than previously thought. Larger multi- are all below detection limits (Fig. 5). ARTICLE IN PRESS 7598 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602

3.5 0.25 concentration dynamics could not be reproduced Predicted formic

d until the reactions GLY+2 OH-2HCOOH and 3.0 Formic Acid, Exp. 1, 2 ) Predicted oxalic 0.20

- 2.5 mM GLY+H2O2 2HCOOH (reactions 17 and 24) Oxalic Acid, Exp. 1, 2 ( were added to the expanded mechanism. The first 2.0 0.15 reaction is reasonable because a-dicarbonyl com- 1.5 0.10 pounds are highly reactive, and the proximity of the 1.0 0.05 Oxalic Acid two carbonyl double bonds enhances their reactivity Formic Acid (mM) 0.5 toward nucleophilic (e.g., dOH) attack (Vollhardt 0.0 0.00 0 100 200 300 400 500 and Schore, 1994). The second reaction route is Time (min.) supported by substantial experimental evidence (Gavagan et al., 1995; Rao and Rao, 2005; Seip et Fig. 7. Measured and predicted concentrations of formic (left al., 1993), although, to the best of our knowledge, a axis) and oxalic (right axis) acids using expanded mechanism (all rate constant for this reaction is not provided in the pathways, Fig. 1; Table 2, reactions 1–28). Glyoxylic acid concentrations are too small (o0.02 mM) to see at this scale. literature. The expanded mechanism therefore Glyoxal is below detection limits by the 9 min sample. includes direct formic acid production from GLY d through reactions with OH and H2O2 (pathway c, Fig. 1). Reactions were added (reactions 21–23) to involving peroxy compounds. Oxalic acid and dOH parameterize the formation and degradation of a (but not H2O2) are sensitive to the choice of k2. class of larger multifunctional intermediates (i.e., Using the expanded mechanism (Fig. 1) and the unresolved carbon) that appear to be respon- reactions 1–28 (Table 2) with certain rate constants sible for most of the oxalic acid formation. These as fitting parameters (k2, k17, k21–k28), predicted and simplified reactions do not explicitly characterize measured concentration profiles are in reasonable the products, as they are unknown and appear to agreement, and the predictions reproduce the change with time. The rate constants (k21–k23) were measured concentration dynamics (Fig. 7). The adjusted so that model predictions fit the observed expanded mechanism describes 81% of the variance concentrations of quantified species (e.g., formic in oxalic acid observations. Formic acid predictions and oxalic acids). Fitting was not directly possible account for 60% of the variance in observations. for the ‘‘higher molecular weight compounds’’ This modest agreement for formic acid might be because they are unidentified and therefore calibra- explained by the lower recovery for formic acid or tion relationships cannot be calculated. However, the fact that formic acid appears very rapidly and the model fit for quantifiable species was excellent degrades early in the experiment, when uncertainty and these calculated rate constants are consistent in the independent variable (i.e., time) is still a with other investigators who note that aqueous- relatively large fraction of the measurement. The phase reaction rate constants involving the hydroxyl predicted glyoxylic acid concentrations using the radical occur at or near diffusion limits (i.e., expanded mechanism are below detection limits, as 1010 M1 s1)(Haag and Yao, 1992; Zhu and observed. The expanded model suggests that less Nicovich, 2003). For both the original mechanism than 1% of oxalic acid in the reaction vessel is and the expanded mechanism, initial concentrations formed through the glyoxylic acid pathway (path- d of H2O2 and OH were optimized at 9 and 0.15 mM, way b, Fig. 1). respectively, reflecting the fact that some H2O2 photolyzes to dOH during solution preparation (see 4.3. Model limitations Supporting Information). Additionally, organic reactions with H2O2 were added to the model Reactions 21–23 are surrogates for poorly under- (reactions 25–28). Including these reactions is stood processes that enhance reactant disappear- reasonable because pyruvic (Stefan and Bolton, ance. For example, GLY could also react with 1999; Carlton et al., 2006) and glyoxylic acids (Seip unidentified products (e.g., radicals or larger multi- et al., 1993; Gavagan et al., 1995; Rao and Rao, functional compounds observed in the HPLC and 2005) have been shown to react with H2O2. Their ESI mass spectra). The larger multifunctional inclusion improved model performance for the products change during the experiment, as indicated experiments and controls. Note that k2 exceeds the by the shifts in the HPLC retention time and as published rate constant for reaction 2 (Table 2) and demonstrated by the distribution of mass species in likely represents the net rate of several reactions the ESI–MS. Further experimental work to identify ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7599 products and their formation mechanisms is re- constants. Plus, while we expect higher molecular quired for further model improvement and for weight material to remain largely in the particle accurate prediction of the properties and concentra- phase upon droplet evaporation (Loeffler et al., tion dynamics of the larger molecular weight 2006), its gas-particle partitioning is not well products. Prediction of SOA formation through characterized. However, this value indicates that aqueous-phase photochemistry is hampered by the model investigations of cloud-produced SOA that lack of understanding of the thermodynamic neglect higher molecular weight compounds (e.g., properties, atmospheric stability and concentrations rely on carboxylic acid production solely) could of these unknown products. substantially under-predict SOA formation.

5. Yields 6. Discussion

In these batch reactor experiments, a fixed This work supports the hypothesis that aqueous- quantity of GLY was reacted with dOH generated phase photooxidation of GLY leads to SOA by H2O2 photolysis leading to the formation, formation. Oxalic acid formation is confirmed, evolution and eventual destruction of low volatility and an expanded formation mechanism is proposed. products. Because the chemical composition in the The coupled atmospheric concentration dynamics reaction vessel continues to evolve after all GLY has of GLY and oxalic acid (Chebbi and Carlier, 1996), reacted (9 min), the mass of product per mass of results from recent field campaigns (Crahan et al., GLY reacted (frequently called the yield) is not a 2004; Sorooshian et al., 2006) and this work support constant, but a function of the reaction time. After the hypothesis that aqueous-phase chemistry is the 10 min of aqueous-phase processing the reaction predominant formation mechanism for oxalic acid. vessel contained 0.01 g oxalic acid per g GLY Good agreement is found between measured and reacted (1%). (Note: 10 min is a typical cloud modeled oxalic acid concentrations using the droplet lifetime and an air parcel might be processed expanded mechanism developed herein. This work through tens of cloud cycles during regional suggests that larger multifunctional compounds are transport; Ervens et al., 2004.) At the oxalic acid also products of aqueous-phase GLY photooxida- concentration maximum (90 min), the reaction tion. Both oxalic acid and larger multi-functional vessel contained 0.02 g oxalic acid per g GLY compounds will contribute SOA upon droplet reacted (2%). After 150 min the mass of oxalic evaporation. acid in the reaction vessel decreased, until all oxalic The larger multifunctional products are likely to acid was oxidized to CO2 (4300 min). Atmospheric be covalently bonded oligomers or compounds with yields of oxalic acid (and SOA) from cloud relatively high pKa values and carboxylic acid or processing of GLY will depend not only on the alcohol functional groups. The evidence supporting aqueous-phase chemistry investigated in this re- this is as follows: the HPLC column retains only search, but also on the rate at which oxidants and carboxylic acids and alcohols, high pKa compounds GLY are (continuously) supplied by gas-phase have longer elution times in this column, and chemistry, on photolysis rates, and on cloud multifunctional compounds have broad chromato- dynamics (e.g., cloud contact time). graphic peaks. Also, the complexity in the ESI mass We expect that the larger multifunctional pro- spectra occurs in both the positive (e.g., alcohols) ducts formed from aqueous GLY photooxidation and negative (e.g., carboxylic acids) modes. The will also contribute to SOA. Experimental observa- formation of larger multi-functional compounds tions suggest that the maximum higher molecular has been reported previously in similar photooxida- weight compound concentration occurs at 30 min, tion experiments of carboxylic acids (e.g., pyruvic when the median molecular weight in the ESI–MS acid; Altieri et al., 2006). These compound classes was 175 g mol1 (Fig. 3). Using this as the (i.e., alcohols, covalently bonded oligomers and estimated molecular weight, a maximum of 0.3 g larger carboxylic acids) are consistent with pro- higher molecular weight compounds g1 reacted posed cloud water HULIS (humic-like substances) GLY (30%) was estimated for the reaction vessel components (Cappiello et al., 2003). Oligomers and (30 min). Note there is substantial uncertainty in HULIS have been found in atmospheric aerosols this estimate as compounds are unknown and there (Gao et al., 2006; Kalberer et al., 2004, 2006) and in is uncertainty in the calculated reaction rate clouds ( Fuzzi et al., 2002; Cappiello et al., 2003). ARTICLE IN PRESS 7600 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602

It has been suggested that these larger compounds of Commerce’s National Oceanic and Atmospheric form in a concentrated liquid phase (i.e., acidic Administration (NOAA) and under agreement particles; Tolocka et al., 2004; Kroll et al., 2006), number DW13921548. This work constitutes a but we provide evidence that larger multi-functional contribution to the NOAA Air Quality Program. compounds can also form in dilute solutions (i.e., Although it has been reviewed by EPA and NOAA cloud droplets) as suggested by Blando and Turpin and approved for publication, it does not necessa- (2000) and Gelencser and Varga (2005). Still, rily reflect their policies or views. further work is needed to better characterize these unidentified products, their formation chemistry Appendix A. Supplementary material and relevant atmospheric properties. The work presented here demonstrates one of the The online version of this article contains addi- many complex pathways leading to the formation of tional supplementary data. Please visit doi:10.1016/ SOA through cloud processing. SOA yields from GLY j.atmosenv.2007.05.035. could be substantially higher than oxalic acid yields because oxalic acid is only one of many potential low volatility products. For example, in addition to oxalic acid and larger multifunctional compounds, organo- sulfur compounds (e.g., GLY–sulfur adducts; Munger References et al., 1984, 1995; Liggio et al., 2005) could form and Allen, A.O., Hochanadel, C.J., Ghormley, J.A., Davis, T.W., contribute to SOA. Also, while GLY is a gas-phase 1952. Decomposition of water and aqueous solutions under oxidation product of many VOCs, it is only one of mixed fast neutron and gamma radiation. Journal of Physical many potential aqueous-phase SOA precursors. Thus, Chemistry 56, 576–586. the inclusion of in-cloud SOA formation pathways in Altieri, K.E., Carlton, A.G., Lim, H.J., Turpin, B.J., Seitzinger, chemical transport models is likely to reduce the gap S., 2006. Evidence for oligomer formations in clouds: reactions of isoprene oxidation products. Environmental between measured and modeled organic PM in the free Science and Technology 40, 4956–4960. troposphere (Heald et al., 2006). Atkinson, R., 2000. Atmospheric chemistry of VOCs and NOx. Atmospheric Environment 34, 2063–2101. Atkinson, R., Baulch, D.L., Cox, R.A., Crowley, J.N., Hampson, Acknowledgments R.F., Hynes, R.G., Jenkin, M.E., Rossi, M.J., Troe, J., 2006. Evaluated kinetic and photochemical data for atmospheric The authors gratefully acknowledge useful conver- chemistry: volume II-gas phase reactions of organic species. sations with Dr. Jeehiun Lee, Dr. John Reinfelder, Yi Atmospheric Chemistry and Physics 6, 3625–4055. Betterton, E.A., Hoffmann, M.R., 1988. Henry’s Law constants Tan, and Dr. Mark Perri. This research was supported of some environmentally important aldehydes. Environmen- in part by the US EPA Science to Achieve Results tal Science and Technology 22, 1415–1418. (STAR) program (R831073), the National Science Blando, J.D., Turpin, B.J., 2000. Secondary organic aerosol Foundation (NSF-ATM-0630298), an Air & Waste formation in cloud and fog droplets: a literature evaluation of Management Association Air Pollution Research plausibility. Atmospheric Environment 34, 1623–1632. Brereton, R.G., 1997. Multilevel multifactor designs for multi- Grant (APERG) and the New Jersey Agricultural variate calibration. Analyst 122, 1521–1529. Experiment Station and the NOAA Climate Goal. Buxton, G.V., Greenstonck, C.L., Hellman, W.P., Ross, A.B., Although the research described in this paper has been 1988. Journal of Physical Chemistry Reference Data 17, 513. Buxton, G.V., Malone, T.N., Salmon, G.A., 1997. Oxidation of funded and reviewed by the US EPA, it does not d necessarily reflect the views of the EPA; no official glyoxal initiated by OH in oxygenated aqueous solution. Journal of the Chemical Society, Faraday Transactions 93, endorsement should be inferred. Any opinions, 2889. findings, and conclusions or recommendations ex- Cappiello, A., De Simoni, E., Fiorucci, C., Mangani, F., Palma, pressed in this material are those of the authors and P., Trufelli, H., Decesari, S., Facchini, M.C., Fuzzi, S., 2003. do not necessarily reflect the views of the National Molecular characterization of the water-soluble organic Science Foundation. compounds in fogwater by ESIMS/MS. Environmental Science and Technology 37, 1229–1240. Carlton, A.G., Lim, H.-J., Altieri, K., Seitinger, S., Turpin, B.J., Disclaimers: The research presented here was 2006. Link Between Isoprene and SOA: fate of pyruvic acid in dilute aqueous solution. Geophysical Research Letters 33. performed, in part, under the Memorandum of Chebbi, A., Carlier, P., 1996. Carboxylic acids in the troposphere, Understanding between the US Environmental occurrence, sources and sinks: a review. Atmospheric Protection Agency (EPA) and the US Department Environment 30, 4233–4249. ARTICLE IN PRESS A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602 7601

Crahan, K.K., Hegg, D., Covert, D.S., Jonsson, H., 2004. An Heald, C.L., Jacob, D.J., Turquety, S., Hudman, R.C., Weber, exploration of aqueous oxalic acid production in the coastal R.J., Sullivan, A.P., Peltier, R.E., Atlas, E.L., de Gouw, J.A., marine atmosphere. Atmospheric Environment 23, Warneke, C., Holloway, J.S., Neuman, J.A., Flocke, F.M., 3757–3764. Seinfeld, J.H., 2006. Concentrations and sources of organic EPA, US, 2004. Air Quality Criteria for Particulate Matter. US carbon aerosols in the free troposphere over North America. Environmental Protection Agency, Research Triangle Park. Journal of Geophysical Research-Atmospheres 111. Elliot, A.J., Buxton, G.V., 1992. Temperature dependences of the Herrmann, H., Reese, A., Zellner, R., 1995. Time-resolved UV/ reactions OH+O2 and HO2 in water upto 200 1C. Journal of Vis diode-array absorption-spectroscopy of SOx-(X ¼ 3, 4, 5) the Chemical Society, Faraday Transactions 88, 2465–2470. radical-anions in aqueous-solution. Journal of Molecular Ervens, B., George, C., Williams, J.E., Buxton, G.V., Salmon, Structure 348, 183–186. G.A., Bydder, M., Wilkinson, F., Dentener, F., Mirabel, P., IPCC, 2001. Climate Change 2001: The Scientific Basis. Cam- Wolke, R., Herrmann, H., 2003a. CAPRAM 2.4 (MODAC bridge University Press, Cambridge, UK. mechanism): an extended and condensed tropospheric aqu- Kalberer, M., Paulsen, D., Sax, M., Steinbacher, M., Dommen, eous phase mechanism and its application. Journal of J., Prevot, A.S.H., Fisseha, R., Weingartner, E., Frankevich, Geophysical Research-Atmospheres 108. V., Zenobi, R., Baltensperger, U., 2004. Identification of Ervens, B., Gligorovski, S., Herrmann, H., 2003b. Temperature- polymers as major components of atmospheric organic dependent rate constants for hydroxyl radical reactions with aerosols. Science 303, 1659–1662. organic compounds in aqueous solutions. Physical Chemistry Kalberer, M., Sax, M., Samburova, V., 2006. Molecular size Chemical Physics 5, 1811–1824. evolution of oligomers in organic aerosols collected in urban Ervens, B., Feingold, G., Frost, G.J., Kreidenweis, S.M., 2004. A atmospheres and generated in a smog chamber. Environ- modeling study of aqueous production of dicarboxylic acids: mental Science & Technology 40, 5917–5922. 1. Chemical pathways and speciated organic mass production. Kanakidou, M., Seinfeld, J.H., Pandis, S.N., Barnes, I., Journal of Geophysical Research 109. Dentener, F.J., Facchini, M.C., Dingenen, R.V., Ervens, B., Fuzzi, S., Facchini, M.C., Decesari, S., Matta, E., Mircea, M., Nenes, A., Nielsen, C.J., Swietlicki, E., Putaud, J.P., 2002. Soluble organic compounds in fog and cloud droplets: Balkanski, Y., Fuzzi, S., Horth, H., Moortgat, G.K., Winterhalter, R., Myhre, C.E.L., Tsigaridis, K., Vignati, E., what have we learned over the past few years? Atmospheric Stephaou, E.G., Wilson, J., 2005. Organic aerosol and global Research 64, 89–98. climate modelling: a review. Atmospheric Chemistry and Gao, S., Surratt, J.D., Knipping, E.M., Edgerton, E.S., Physics 5, 1053–1123. Shahgholi, M., Seinfeld, J.H., 2006. Characterization of polar Kawamura, K., Gagosian, R.B., 1987. Implication of o- organic components in fine aerosols in the southeastern oxocarboxylic acids in the remote marine atmosphere for United States: identity, origin, and evolution. Journal of photo-oxidation of unsaturated fatty acids. Nature 325, Geophysical Research-Atmospheres 111. 330–332. Gavagan, J.E., Fager, S.K., Seip, J.E., Payne, M.S., Anton, D.L., Kawamura, K., Usukura, K., 1993. Distributions of low DiComsimo, R., 1995. Glyoxylic acid production using molecular weight dicarboxylic acids in North Pacific aerosol microbial transformant catalysts. Journal of Organic Chem- samples. Journal of Oceanography 49, 271–283. istry 60, 3957–3963. Kawamura, K., Kasukabe, H., Barrie, L., 1996a. Source and Gelencser, A., Varga, Z., 2005. Evaluation of the atmospheric reaction pathways of dicarboxylic acids, ketoacids, and significance of multiphase reactions in atmospheric secondary dicarbonyls in Arctic aerosols: one year of observations. organic aerosol formation. Atmospheric Chemistry and Atmospheric Environment 30, 1709–1722. Physics 5, 2823–2831. Kawamura, K., Semere, R., Imai, Y., Fujii, Y., Hayashi, M., Greenberg, A.E., Eaton, A.D., Clesceri, L.S., 1991. American 1996b. Water soluble dicarboxylic acids and related com- Public Health Association, American Water Works Associa- pounds in Antarctic aerosols. Journal of Geophysical tion, and Water Pollution Control Federation, Standard Research-Atmospheres 101, 18721–18728. Methods for the Examination of Water and Wastewater: 1991 Klassen, N.V., Marchinton, D., McGowan, H.C.E., 1994. H2O2 Supplement to the 17th Edition. American Public Health determination by the I3 method and by KMNO4 titration. Association, Washington, DC, 1449 pp. Analytical Chemistry 66, 2921–2925. Guzman, M.I., Colussi, A.I., Hoffmann, M.R., 2006. Photo- Kleindienst, T.E., Smith, D.F., Li, W., Edney, E.O., Driscoll, induced oligomerization of aqueous pyruvic acid. Journal of D.J., Speer, R.E., Weathers, W.S., 1999. Secondary organic Physical Chemistry A 110, 3619–3626. aerosol formation from the oxidation of aromatic hydro- Haag, W.R., Yao, C.C.D., 1992. Rate constants for reaction of carbons in the presence of dry submicron ammonium sulfate hydroxyl radicals with several drinking water contaminants. aerosol. Atmospheric Environment 33, 3669–3681. Environmental Science & Technology 26, 1005–1013. Kroll, J.H., Ng, N.L., Murphy, S.M., Flagan, R.C., Seinfeld, Hastings, W.P., Koehler, C.A., Bailey, E.L., DeHaan, D.O., J.H., 2006. Secondary organic aerosol formation from 2005. Secondary organic aerosol formation by glyoxal isoprene photooxidation. Environmental Science and Tech- hydration and oligomer formation: humidity effects and nology 40, 1869–1877. equilibrium shifts during analysis. Environmental Science and Kunen, S.M., Lazrus, A.L., Kok, G.L., Heikes, B.G., 1983.

Technology 39, 8728–8735. Aqueous oxidation of SO2 by hydrogen peroxide. Journal of Heald, C.L., Jacob, D.J., Park, R.J., Russell, L.M., Huebert, Geophysical Research 88, 3671–3674. B.J., Seinfeld, J.H., Liao, H., Weber, R.J., 2005. A large Lelieveld, J., Crutzen, P.J., 1991. The role of clouds in organic aerosol source in the free troposphere missing from tropospheric photochemistry. Journal of Atmospheric Chem- current models. Geophysical Research Letters 32. istry 12, 229. ARTICLE IN PRESS 7602 A.G. Carlton et al. / Atmospheric Environment 41 (2007) 7588–7602

Liao, C.-H., Gurol, M.D., 1995. Chemical oxidation by photo- organic matter in stream water using electrospray-ioniation lytic decomposition of hydrogen peroxide. Environmental mass spectrometry. Limnology and Oceanography 50, 1–12. Science and Technology 29. Sempere, R., Kawamura, K., 1994. Comparative distributions of Liggio, J., Li, S.M., McLaren, R., 2005. Heterogeneous reactions dicarboxylic acids and related polar compounds in snow, rain of glyoxal on particulate matter: identification of acetals and and aerosols from urban atmosphere. Atmospheric Environ- sulfate esters. Environmental Science and Technology 39, ment 28, 449–459. 1532–1541. Sharp, J.H., Benner, R., Bennett, L., Carlson, C.A., Dow, R., Lim, H.-J., Carlton, A.G., Turpin, B.J., 2005. Isoprene forms Fitzwater, S.E., 1993. Reevaluation of high temperature secondary organic aerosol through cloud processing: model combustion and chemical oxidation measurements of dis- simulations. Environmental Science and Technology 39, solved organic-carbon in seawater. Limnology and Oceano- 4441–4446. graphy 38, 1774. Loeffler, K.W., Koehler, C.A., Paul, N.M., De Haan, D.O., 2006. Sorooshian, A., Varutbangkul, V., Brechtel, F.J., Ervens, B., Oligomer formation in evaporating aqueous glyoxal and Feingold, G., Bahreini, R., Murphy, S.M., Holloway, J.S., methyl glyoxal solutions. Environmental Science and Tech- Atlas, E.L., Buzorius, G., Jonsson, H., Flagan, R.C., Seinfeld, nology 40, 6318–6323. J.H., 2006. Oxalic acid in clear and cloudy atmospheres: Magneron, I., Mellouki, A., LeBras, G., Moortgat, G.K., analysis of data from ICARTT 2004. Journal of Geophysical Horowitz, A., Wirtz, K., 2005. Photolysis and OH-initiated Research-Atmospheres 111, D23S45. oxidation of glycolaldehyde under atmospheric conditions. Spaulding, R.S., Schade, G.W., Goldstein, A.H., Charles, M.J., Journal of Physical Chemistry A 109, 4552–4561. 2003. Characterization of secondary atmospheric photoox- Martens, H., Naes, T., 1989. Multivariate Calibration. Wiley, idation products: evidence for biogenic and anthropogenic New York. sources. Journal of Geophysical Research-Atmospheres 108. Matsumoto, K., Kawai, S., Igawa, M., 2005. Dominant factors Stefan, M.I., Bolton, J.R., 1999. Reinvestigation of the acetone controlling concentrations of aldehydes in rain, fog, dew degradation mechanism in dilute aqueous solution by the UV/ water, and in the gas phase. Atmospheric Environment 39, H2O2 process. Environmental Science and Technology 30, 7321–7329. 870–873. Miller, W.L., Moran, M.A., 1997. Interaction of photochemical Stefan, M.I., Hoy, A.R., Bolton, J.R., 1996. Kinetics and and microbial processes in the degradation of refractory mechanism of the degradation and mineralization of acetone dissolved organic matter from a coastal marine environment. in dilute aqueous solution sensitized by the UV photolysis of Limnology and Oceanography 42, 1317–1324. hydrogen peroxide. Environmental Science and Technology Munger, J.W., Jacob, D.J., Hoffmann, M.R., 1984. The 30, 2382–2390. occurrence of bisulfite-aldehyde addition products in fog- Tolocka, M.P., Jang, M., Ginter, J.M., Cox, F.J., Kamens, R.M., and cloud-water. Journal of Atmospheric Chemistry 1, Johnston, M.V., 2004. Formation of oligomers in secondary 335–350. organic aerosol. Environmental Science and Technology 38, Munger, J.W., Collett, J., Daube Jr., B.C., Hoffmann, M.R., 1428–1434. 1990. Fogwater chemistry at Riverside California. Atmo- Volkamer, R., Platt, U., Wirtz, K., 2001. Primary and secondary spheric Environment 24, 185–205. glyoxal formation from aromatics: experimental evidence for Munger, J.W., Jacob, D.J., Daube, B.C., Horowitz, L.W., 1995. the bicycloalkyl-radical pathway from benzene, toluene, and Formaldehyde, glyoxal and methylglyoxal in air and cloud- water at a rural mountain site in central Virginia. Journal of p-xylene. Journal of Physical Chemistry A 105, 7865–7874. Geophysical Research 100, 9325–9333. Volkamer, R., Molina, L.T., Molina, M.J., Shirley, T., Brune, Murov, S.L., 1973. Handbook of Photochemistry. Marcel W.H., 2005. DOAS measurements of glyoxal as an indicator Decker, Inc., New York. for fast VOC chemistry in urban air. Geophysical Research Rahn, R.O., Stefan, M.I., Bolton, J.R., Goren, E., Shaw, P.S., Letters 32 Art. No. L08806. Lykke, K.R., 2003. Quantum yield of the iodide–iodate Vollhardt, K.P.C., Schore, N.E., 1994. Organic Chemistry, chemical actinometer: dependence on wavelength and con- second ed. W.H. Freeman and Company, New York. centration. Photochemistry and Photobiology 78, 146–152. Warneck, P., 1999. The relative importance of various pathways Rao, T.S., Rao, M.S.P., 2005. Titrimetric and spectrophoto- for the oxidation of sulfur dioxide and nitrogen dioxide in metric methods for the determination of glyoxal and analysis sunlit continental fair weather clouds. Physical Chemistry of ternary mixtures of its oxidation products. Journal of Chemical Physics 1, 5471. Analytical Chemistry 60, 714–718. Warneck, P., 2003. In-cloud chemistry opens pathway to the Schweitzer, F., Magi, L., Mirabel, P., George, C., 1998. Uptake formation of oxalic acid in the marine atmosphere. Atmo- rate measurements of methanesulfonic acid and glyoxal by spheric Environment 37, 2423–2427. aqueous droplets. Journal of Physical Chemistry A 102, Whipple, E.B., 1970. The structure of glyoxal in water. Journal of 593–600. the American Chemical Society 92, 7183–7186. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Yu, J.Z., Huang, X.-F., Xu, J., Hu, M., 2005. When aerosol Physics. Wiley, New York. sulfate goes up, so does oxalate: implication for the formation Seip, J.E., Fager, S.K., Gavagan, J.E., Gosser, L.W., Anton, mechanisms of oxalate. Environmental Science and Technol- D.L., DiCosimo, R., 1993. Biocatalytic production of ogy 29, 128–133. glyoxylic acid. Journal of Organic Chemistry 58, 2253–2259. Zhu, L., Nicovich, J.M., et al., 2003. Temperature-dependent Seitzinger, S.P., Hartnett, H., Lauck, R., Mazurek, M., Mine- kinetics studies of aqueous phase reactions of hydroxyl gishi, T., Spyres, G., Styles, R., 2005. Molecular-level radicals with dimethylsulfoxide, dimethylsulfone, and metha- chemical characterization and bioavailability of dissolved nesulfonate. Aquatic Sciences 65 (4), 425–435.